Implementing defeasible reasoning